Tonic and Spillover Inhibition of Granule Cells Control Information Flow Through Cerebellar Cortex

Home , Bicuculline, Furosemide, Pregnenolone sulfate

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector Neuron, Vol. 33, 625–633, February 14, 2002, Copyright 2002 by Cell Press Tonic and Spillover Inhibition of Granule Cells Control Information Flow through Cerebellar Cortex

Martine Hamann, David J. Rossi, the fraction of granule cells excited by mossy fiber input and David Attwell1 to the cerebellum and, thus, to increase cerebellar infor- Department of Physiology mation storage capacity (Marr, 1969; Tyrrell and Will- University College London shaw, 1992). However, the relative importance of tonic Gower Street versus mossy fiber driven inhibition and their effect on London, WC1E 6BT the input-output relationship of the cerebellar cortex United Kingdom have not been studied, largely because until recently, it was not possible to block inhibition of granule cells without also blocking inhibition of Purkinje cells. Summary Here, we show that both spillover and tonic inhibition of adult granule cells are mediated largely by receptors We show that information flow through the adult cere- that are blocked by furosemide and insensitive to diaze- bellar cortex, from the mossy fiber input to the Purkinje pam and neurosteroids. We quantify the relative impor- cell output, is controlled by furosemide-sensitive, di- tance of tonic inhibition and of the direct and spillover

azepam- and neurosteroid-insensitive GABAA recep- components of synaptically evoked inhibition. Finally, tors on granule cells, which are activated both tonically we demonstrate that tonic and spillover inhibition of and by GABA spillover from synaptic release sites. granule cells play a major role in controlling information Tonic activation of these receptors contributes a flow through the cerebellar cortex. 3-fold larger mean inhibitory conductance than GABA released synaptically by high-frequency stimulation. Results Tonic and spillover inhibition reduce the fraction of granule cells activated by mossy fiber input, generat- Tonic Inhibition of Granule Cells ing an increase of coding sparseness, which is pre- In cerebellar slices from adult rats, applying the GABAA dicted to improve the information storage capacity of receptor blocker bicuculline to granule cells whole-cell the cerebellum. clamped at Ϫ60 mV with electrodes containing isotonic Ϫ Cl (ECl ϭ 0 mV) produced an outward current shift of Introduction 42 Ϯ 4 pA, reflecting the suppression of a tonic conductance of 700 Ϯ 60 pS generated by GABAA receptors Transmitter spilling out of the synaptic cleft can generate (Figure 1A). Similar results were seen with the GABAA ␮ a current in distant neurons if their receptors have a blockers GABAzine (Figure 1G) and picrotoxin (100 M, sufficiently high affinity (EC Ϸ 1 ␮M) to respond to two cells; data not shown). The current was not reduced 50 Ϯ ϭ ␮ the low concentration of transmitter reaching them; for (0.6% 2.9%, p 0.85) by TTX (1 M, Figures 1A and example, NMDA receptors in the case of glutamate 1C; cf. Kaneda et al., 1995; Wall and Usowicz, 1997) or 2ϩ Ϯ ϭ (Isaacson, 1999) or GABA receptors containing the ␣ by removing external Ca (9.6% 3.8%, p 0.09, A 6 2ϩ subunit (Rossi and Hamann, 1998). High-affinity recep- replaced by Mg and 2 mM EGTA; Figures 1B and 1C). Thus, the tonic conductance is not activated by GABA tors may also generate a current even in the absence released by action potentials or Ca2ϩ influx. of synaptic transmitter release if the background transmitter concentration maintained by transporters is in the Tonic Inhibition Is Mediated by Furosemide- low micromolar range that activates the receptors. Tonic Sensitive, Diazepam- and Neurosteroid- activation by low background levels of transmitter may Insensitive GABA Receptors occur for hippocampal and cerebellar NMDA receptors A In adult granule cells, most GABA receptors contain ␣ (Sah et al., 1989; Rossi and Slater, 1993) and in adult A 1 or ␣ subunits together with ␤ or ␤ and ␥ or ␦ subunits cerebellar granule cells that show an action potential- 6 2 3 2 (Laurie et al., 1992a; Quirk et al., 1994; Wisden et al., independent tonic GABA receptor conductance (Kaneda A 1996; Nusser et al., 1998, 1999; Pirker et al., 2000). Furo- et al., 1995; Tia et al., 1996a; Brickley et al., 1996; Wall semide (100 ␮M) inhibits receptors containing ␣6 sub- and Usowicz, 1997). units (with an IC of 10–40 ␮M), but not ␣ -containing In cerebellar granule cells, the molecular identity, cel- 50 1 receptors lacking ␣6 subunits (IC50 Ͼ 3 mM; Korpi et al., lular location, and functional role of the GABAA receptors 1995; Sigel and Baur, 2000), irrespective of the presence mediating spillover and tonic inhibition are uncertain. of ␥2 or ␦ subunits (Korpi and Lu¨ ddens, 1997; Thomp- Pharmacology shows that the receptors mediating spill- son et al., 1999). Furosemide produced an outward cur- ␣ over inhibition contain 6 subunits (Rossi and Hamann, rent (Figures 1D and 1G) that was 58% Ϯ 4% of the 1998), but the receptors’ other subunits are unknown. current produced by bicuculline in the same cells. No ␣ Mice lacking 6 subunits lack the tonic conductance current was produced by furosemide in bicuculline (Fig- (Brickley et al., 2001), but this need not imply that tonic ures 1D and 1G). Furosemide did not generate a current ␣ inhibition is due to receptors containing 6 subunits be- by blocking KCl cotransporters (Payne, 1997) and alter- cause of side effects of the ␣ knockout (see Discussion). Ϫ 6 ing [Cl ]i since the KCl cotransport blocker bumetanide Inhibition of granule cells has been postulated to reduce did not affect the membrane current (Figures 1E and 1G) when applied at a dose (220 ␮M) producing the as 100 ␮M (%80ف) 1Correspondence: [email protected] same inhibition of KCl transporters Neuron 626

Figure 1. Tonic Inhibition of Adult Granule Cells Is by Furosemide-Sensitive, Diazepam-

Insensitive GABAA Receptors (A) Response of a granule cell at Ϫ60 mV

(ECl ϭ 0 mV, so GABAA currents are inward) to bicuculline (40 ␮M) and to bicuculline in the presence of TTX (1 ␮M). Dotted lines show zero current. (B) Response to bicuculline in normal solution and in zero Ca2ϩ solution. (C) Mean bicuculline-evoked currents in the presence and absence of TTX (four cells) and of Ca2ϩ (four cells). (D) Response to furosemide (100 ␮M) and to furosemide in the presence of bicuculline (40 ␮M). (E) Response to furosemide and to bumetanide (220 ␮M). (F) Response to diazepam (0.5 ␮M) and to bicuculline. (G) Mean current shifts at Ϫ60 mV evoked by bicuculline (bic; 40 ␮M, 45 cells), GABAzine (10 ␮M, n ϭ 9), furosemide (furo; 100 ␮M, n ϭ 18), furosemide in bicuculline (n ϭ 4), bumetanide (bum; 220 ␮M, n ϭ 12), and diazepam (dz; 0.5 ␮M, n ϭ 7). Traces are filtered at 2 Hz, so individual spontaneous IPSCs are not visible.

furosemide (Payne, 1997). Furthermore, as shown be- (Zhu et al., 1996; Wohlfarth et al., 2002; see Discussion). low, furosemide did not affect the holding current of In adult granule cells, the neurosteroid THDOC (3␣,21- Golgi, Purkinje, stellate, or basket cells that do not ex- dihydroxy-5␣-pregnan-20-1, 0.1 ␮M) generated an in- press ␣6 subunits (Laurie et al., 1992a), nor did it generate ward current (11 Ϯ 3 pA) corresponding to a 34% Ϯ 5% a current in granule cells from 6-day-old rats (n ϭ 5, potentiation of the bicuculline-sensitive tonic current in data not shown), at which age there is little expression the same cell (Figures 2A and 2C), which was abolished of ␣6 subunits (Laurie et al., 1992b). Diazepam (0.5 ␮M), by bicuculline (40 ␮M, n ϭ 4) and reduced 43% Ϯ 10% which potentiates 2- to 3-fold the response of ␣1-con- by furosemide (100 ␮M, n ϭ 5; data not shown). By taining receptors, but not of ␣6-containing receptors contrast, in granule cells in 6-day-old rats that do not (Saxena and MacDonald, 1996; Sigel and Baur, 2000), yet express ␦ subunits (Laurie et al., 1992b) and show did not affect the tonic current in adult granule cells no tonic current (Brickley et al., 1996; Wall and Usowicz, (Figures 1F and 1G), although it did potentiate (nonspillo- 1997), THDOC potentiated the response to 5 ␮M exoge- ver) synaptic responses mediated by ␣1-containing re- nous GABA by 84% Ϯ 15% (Figures 2B and 2C). Simi- ceptors in these cells (Figure 2H). Thus, the tonic current larly, pregnenolone sulfate (1 ␮M) did not affect the tonic in mature granule cells is generated exclusively by furo- GABAA current of adult granule cells (Figures 2D and semide-sensitive, diazepam-insensitive GABAA recep- 2F) but did inhibit the response of 6-day-old granule tors; i.e., receptors containing ␣6 subunits (see Dis- cells to 5 ␮M GABA (Figures 2E and 2F). Thus, compared cussion). to GABA-evoked currents in young animals, the tonic

In granule cells, ␣6 subunits coassemble with ␥2 and/or current in adult granule cells is insensitive to neuroste- ␦ subunits (Nusser et al., 1998; Quirk et al., 1994; Jech- roids. The likely receptor subunit composition conferring linger et al., 1998). Neurosteroids modulate GABAA re- this pharmacological profile (furosemide-sensitive, diaz- ceptors, but, in receptors containing ␣6 or ␣1 subunits, epam- and neurosteroid-insensitive) is assessed in the this modulation is altered by the presence of ␦ subunits Discussion section. Tonic and Spillover Inhibition 627

Figure 2. Inhibition of Granule Cells by Neu-

rosteroid-Insensitive GABAA Receptors (A) Potentiation of the tonic current in an adult (P35) granule cell by THDOC (0.1 ␮M); response to bicuculline shows size of tonic cur-

rent; ECl ϭ 0 mV. (B) Potentiation by THDOC of the response of a young (P6) granule cell to GABA (5 ␮M). (C) Mean potentiation of P35 tonic current (eight cells) and P6 GABA response (six cells) by THDOC. (D) Pregnenolone sulfate (PS; 1 ␮M) does not alter the tonic current in P35 granule cells. (E) PS reduces the response of P6 granule cells to GABA (5 ␮M). (F) Mean change by PS of P35 tonic current (eight cells) and P6 GABA response (seven cells). (G) Golgi cell-granule cell IPSC (noisy trace, evoked at 0.3 Hz) fitted with the sum of two decaying exponentials (smooth lines; ␶ϭ29 and 767 ms). (H) Effect of diazepam (0.5 ␮M) on the IPSC (lack of effect on peak, as in Rossi and Ha- mann [1998], is consistent with an increased GABA affinity and receptor saturation at the peak; Otis and Mody, [1992]). (I) Effect of furosemide (100 ␮M) on the IPSC. (J) Effect of THDOC and PS on the spillover component of the IPSC (data enlarged from the time of the square in [I]). (K) Percentage change of the early and spillover IPSC components (30 and 200 ms after stimulus, see Experimental Procedures) produced by diazepam (0.5 ␮M, five cells), furosemide (100 ␮M, five cells), THDOC (0.1 ␮M, seven cells), and PS (1 ␮M, four cells). (L) Charge transfer by two components of IPSC (see Experimental Procedures) in seven cells. All cells at Ϫ60 mV.

The Golgi-Granule Cell IPSC Is Largely Mediated pam- and neurosteroid-insensitive receptors. As in by Furosemide-Sensitive, Diazepam- and young animals (Rossi and Hamann, 1998), the spillover Neurosteroid-Insensitive Receptors component contributed the majority of the IPSC charge Granule cells in adult cerebellum, like those in young transfer (89% Ϯ 3% in seven cells; Figures 2G and 2L). animals (Rossi and Hamann, 1998), show a biphasic IPSC evoked by Golgi cell stimulation (Figure 2G). This has a rapidly decaying component that is potentiated The Effect of Tonic Inhibition on Granule by diazepam and is thus generated partly by ␣1 subunit- Cell Excitability containing receptors, followed by a slow component To test the effect of the furosemide-sensitive, diazepam- mediated by spillover of GABA released from Golgi cells and neurosteroid-insensitive tonic conductance on the that are not anatomically presynaptic to the granule cell excitability of adult granule cells, we applied furosemide (Rossi and Hamann, 1998), which is unaffected by diaze- to cells current clamped using a pipette solution provid- Ϫ pam and, so, is generated by receptors lacking ␣1 sub- ing a physiological Cl gradient (ECl ϭϪ61 mV; Brickley units (Figures 2H and 2K). Furosemide reduced both the et al., 1996). Granule cells had a resting membrane po- early and late components of the IPSC in adult granule tential, input resistance, and capacitance of Ϫ69.6 Ϯ cells (Figures 2I and 2K) implying that ␣6 subunit-con- 2.0 mV, 3.4 Ϯ 0.2 G⍀, and 3.9 Ϯ 0.2 pF (n ϭ 13), respec- taining receptors contribute to both phases. The spill- tively. In response to current injection, they generated over component was essentially unaffected by THDOC trains of action potentials (Figure 3A), the frequency and pregnenolone sulfate (potentiated by 5.7% Ϯ 6.2% of which increased with the injected current amplitude and 2.4% Ϯ 9.0%, respectively; Figures 2J and 2K) in (Figures 3A and 3B). Furosemide (100 ␮M) increased contrast to the early phase (potentiated by 74.3% Ϯ the input resistance and time constant (by 19% Ϯ 4%, 15.5% and inhibited by 11.8% Ϯ 3.5%, respectively; n ϭ 13, p ϭ 0.006), lowered the threshold for action Figure 2K). Thus, as for the tonic current, the spillover potential initiation from 8.2 Ϯ 1.0 to 5.0 Ϯ 0.9 pA (p ϭ component is mediated by furosemide-sensitive, diaze- 0.003, n ϭ 9), and increased the number of action poten- Neuron 628

creased the number of action potentials elicited by a single shock in these cells (Figure 5A). In vivo mossy fibers fire at high frequencies, so we delivered trains of stimuli to the white matter (Figure 5B). The action potential frequency evoked in granule cells increased with, but was always less than, the frequency of mossy fiber stimulation. Furosemide (100 ␮M) approximately doubled the granule cell action potential frequency at all stimulation frequencies (Figures 5C and 5D). To determine whether this was due more to suppression of the tonic conductance or of the synaptically evoked inhibitory conductance, we compared the charge movement per 200 ms suppressed by furosemide in the absence of stimulation (the tonic current) with the suppression by furosemide of charge entry evoked by stimulation, in granule cells voltage-clamped with electrodes containing isotonic ClϪ. At all stimulation frequencies, the furosemide-sensitive, stimulus-evoked charge entry was less than 15% of the furosemide-sensitive tonic charge entry (Figures 5E and 5F). Similarly, the stimulus-evoked bicuculline-sensitive charge movement was less than 32% of the bicuculline-sensitive tonic charge entry (Figures 5E and 5G). Thus, the tonic conductance is the main inhibitory influence on the granule cells: even at a stimulation frequency of 200 Hz, Golgi cell IPSCs generate a mean conductance that is only 1/3 of the tonic conductance or 1/4 of the total inhibitory conductance (and this fraction will be lower at lower stimulation frequencies; Figure 5G), so the tonic conductance provides 3/4 of the inhibition. ␣ Figure 3. Blocking 6 Subunit-Containing GABAA Receptors In- Since the spillover component of the IPSC provides creases Granule Cell Excitability 89% of the IPSC charge transfer (Figure 2L), it follows Ϫ (A) Response of a granule cell (at 65 mV) to current injection before that furosemide-sensitive, diazepam- and neurosteroid- (control), during, and after (wash) application of 100 ␮M furosemide. insensitive receptors provide a total of 89% of 1/4 (for Dotted line shows 0 mV; ECl ϭϪ61 mV. (B) Granule cell firing rate (averaged over 100 ms current pulse) as the spillover IPSC) plus 3/4 (for the tonic conductance) a function of injected current for cell of (A). or 97% of the total inhibitory charge transfer (and this (C) Firing rate as a function of injected current in nine cells in the percentage will be higher at lower stimulation frequen- absence and presence of furosemide (normalized to the rate pro- cies). The fact that furosemide-sensitive, diazepam- and duced by 20 pA in control solution). neurosteroid-insensitive receptors totally dominate inhibition of granule cells implies that the effects of furosemide on mossy fiber to granule cell transmission de- tials elicited at all levels of current injection (Figures scribed above, and on information flow from mossy 3A–3C). fibers to Purkinje cells described in the next section, By contrast, in Golgi cells (Figures 4A–4C; three cells) are almost entirely due to blockage of these receptors. and Purkinje cells (Figures 4D–4F; six cells), furosemide This is because less than 3% of the inhibitory charge did not affect the input resistance, action potential transfer into granule cells is via the early IPSC compo- threshold, or the number or frequency of action poten- nent mediated by neurosteroid-sensitive receptors (and tials elicited by current injection, consistent with a lack not even all of these will contain the ␣ subunits neces- ␣ 6 of expression of 6 subunit-containing receptors (Laurie sary to be inhibited by furosemide). et al., 1992a). Similarly, in 12 interneurons located at various levels in the molecular layer, some of which were The Effect of Tonic and Spillover Inhibition filled with Lucifer yellow and identified as stellate cells on Information Flow through (Figure 4G; five cells) or basket cells (Figure 4H; two the Cerebellar Cortex cells), furosemide did not affect the input resistance. Granule cell axons not only excite Purkinje cells, but also inhibitory Golgi, stellate, and basket cells. In turn, The Effect of Furosemide-Sensitive, stellate and basket cells inhibit Purkinje cells, and Golgi Diazepam- and Neurosteroid-Insensitive cells inhibit granule cells, so it is unclear whether in- Inhibition on Mossy Fiber to Granule Cell creasing the granule cell sensitivity to mossy fiber input Transmission by blocking tonic and spillover inhibition will increase A single electrical shock to the white matter elicited a or decrease the Purkinje cell output. We made current glutamate receptor-mediated mossy fiber EPSP in gran- clamp recordings of Purkinje cell voltage responses to ule cells that elicited an action potential in two out of electrical stimulation of the mossy fibers (with ECl ϭϪ61 four cells tested (Figure 5A). Furosemide (100 ␮M) in- mV). Above 20 Hz stimulation frequency, summating Tonic and Spillover Inhibition 629

Figure 4. Blocking ␣6 Subunit-Containing

GABAA Receptors Has No Effect on Golgi, Purkinje, Stellate, or Basket Cells (A) Response of a Golgi cell (at Ϫ65 mV) to current injection in control solution and in 100

␮M furosemide (ECl ϭϪ61 mV). (B) Firing rate (averaged over 100 ms current pulse) as a function of injected current for the cell of A (typical of three cells). (C) Furosemide (100 ␮M) does not affect the holding current at Ϫ60 mV of a Golgi cell,

although GABA evoked a large current (ECl ϭ 0 mV). (D) Response of a Purkinje cell (at Ϫ70 mV) to current injection before and during su-

perfusion of 100 ␮M furosemide (ECl ϭ Ϫ61 mV). (E) Firing rate (averaged over 200 ms current pulse) as a function of injected current for six cells as in (D). (F) Furosemide (100 ␮M) does not affect the holding current at Ϫ60 mV of a Purkinje cell (done in three cells), although GABA evoked

a large current (ECl ϭ 0 mV). (G) Furosemide (100 ␮M) does not affect the holding current at Ϫ60 mV of a stellate cell,

although GABA evoked a large current (ECl ϭ 0 mV). (H) Furosemide (100 ␮M) does not affect the holding current at Ϫ60 mV of a basket cell,

although GABA evoked a large current (ECl ϭ 0 mV).

EPSPs consistently elicited action potentials, the fre- pected if furosemide acts solely on ␣6 subunit-con- quency of which increased with the stimulation fre- taining GABA receptors in granule cells. quency (Figures 6A, 6B, and 6F). Furosemide (100 ␮M) Thus, by reducing tonic and IPSC inhibition in granule potentiated the EPSPs (Figures 6C and 6D), which re- cells, mediated almost entirely by furosemide-sensitive, duced the time to the first action potential (Figure 6E) diazepam- and neurosteroid-insensitive receptors, furo- and increased the frequency of action potentials at all semide potentiates signal transmission through the cer- stimulation frequencies (Figures 6A, 6B, and 6F). The ebellar cortex. KCl cotransport blocker bumetanide did not affect the response to mossy fiber stimulation (Figure 6G; n ϭ 4). Discussion These experiments were carried out in parasagittal slices in which granule cell to Purkinje cell synapses In this paper, we have shown that the tonic GABAA re- may be mainly made by the ascending granule cell axon. ceptor-mediated inhibitory conductance and the spillo- In case this altered the balance of excitation and inhibi- ver component of Golgi cell IPSCs in adult cerebellar tion received by the Purkinje cells compared to the in granule cells are generated by furosemide-sensitive, di- vivo situation, we also performed experiments on coro- azepam- and neurosteroid-insensitive receptors. Our nal slices, with the parallel fiber part of the axon intact. quantification of the relative importance for granule cells These showed a similar furosemide-evoked potentiation of tonic inhibition and of the direct and spillover compo- of the Purkinje cell response to mossy fiber stimulation nents of synaptic inhibition demonstrates the impor- (Figure 6H; seven cells). However, furosemide had no tance of these receptors. Even when the mossy fibers effect on signal transmission from parallel fibers to Pur- are stimulated at 200 Hz, tonic inhibition provides 3-fold kinje cells in coronal slices (Figure 6I; five cells), as ex- more inhibitory charge transfer than do IPSCs from Golgi Neuron 630

Figure 5. Blocking ␣6 Subunit-Containing

GABAA Receptors Potentiates Signal Trans- mission from Mossy Fibers to Granule Cells Mainly by Suppressing the Tonic Conduc- tance (A) Response of a granule cell (at Ϫ65 mV) to single stimulation of the mossy fibers (arrows) before, during, and after application of 100

␮M furosemide (ECl ϭϪ61 mV). (B) As (A), but 10 Hz stimulation. Right panel shows block of the response by the glutamate receptor blockers NBQX (20 ␮M) and AP5 (50 ␮M). (C) Mean granule cell firing rate in four cells (averaged over the stimulus train: 200 ms for frequencies Ն 20 Hz, 400 ms for 10 Hz) as a function of the mossy fiber (MF) stimulation frequency in the absence and presence of furosemide. (D) As (C), but normalized to the firing rate evoked by 100 Hz stimulation in control con- ditions. (E) Response of a granule cell voltage-

clamped at Ϫ60 mV (ECl ϭ 0 mV) to MF stimulation (black dots; top trace: one stimulus; middle trace: 50 Hz; bottom trace: 200 Hz) in control solution (gray traces) and bicuculline (40 ␮M; top two black traces; remaining inward currents are EPSCs) or furosemide (100 ␮M; bottom black trace). Bicuculline/furosemide-evoked change of baseline current has been subtracted. (F) Furosemide-sensitive charge entry per 200 ms carried by tonic current (as in Figure 1D) and by inhibitory synaptic current (difference of area of traces in [E]) at different stimulation rates in seven cells. (G) As (F), but for bicuculline (40 ␮M, six cells). (F) and (G) differ because 100 ␮M furosemide

blocks ␣6-containing receptors incompletely and because some of the IPSC is mediated

by receptors lacking ␣6 subunits.

cells (Figure 5G), and the spillover component makes up Interpretation of the insensitivity of the tonic and spill- 89% of the IPSC (Figure 2L). Consequently, furosemide- over currents to neurosteroids is complicated by incon- sensitive, diazepam- and neurosteroid-insensitive re- sistencies in the literature. In HEK cells, the presence ceptors mediate at least 97% of the inhibition of granule of the ␦ subunit is reported to greatly decrease the sensi- cells and, as described below, have a profound effect tivity of ␣6 subunit-containing receptors to neurosteroids on information flow through the cerebellar cortex. (Zhu et al., 1996): THDOC (0.1 ␮M) doubles the response

What subunit composition is expected to confer sen- of ␣6␤3 and ␣6␤3␥2 receptors to GABA but potentiates sitivity to furosemide and insensitivity to diazepam and ␣6␤3␥2␦ and ␣6␤3␦ receptor responses by only 34% and neurosteroids on GABAA receptors? In adult granule 0%, respectively (Zhu et al., 1996), while pregnenolone cells, most GABAA receptors contain ␣1 and/or ␣6 sub- sulfate (1 ␮M) reduces the response of ␣6␤3␥2 receptors units (Laurie et al., 1992a; Quirk et al., 1994; Wisden et (by 37%), but not of ␣6␤3␦ receptors (13%, not signifi- al., 1996; Nusser et al., 1998, 1999; Pirker et al., 2000). cant). Consistent with this, Zhu et al. (1996) also demon- The tonic current and the spillover component of the strated that insensitivity of granule cell GABA receptors IPSC are sensitive to furosemide, but not to diazepam. to THDOC was correlated with the expression of ␦ sub-

The IC50 for furosemide blocking receptors containing unit mRNA, and we found that responses to these ste-

␣6, ␣6 and ␣1,or␣1 subunits is 10–40, 300, or 3000 ␮M, roids were much larger early in development (Figures respectively, while diazepam (0.5 ␮M) potentiates the 2C and 2F; see also Zhu and Vicini, 1997) before much response of these receptors by 0%, 50%, and 100%– ␦ subunit is expressed (Laurie et al., 1992b). Similarly, 200%, respectively (Korpi et al., 1995; Saxena and Mac- the early component of the IPSC, which is thought to Donald, 1996; Sigel and Baur, 2000). Thus, the block by be generated by synaptic receptors (Rossi and Hamann, 100 ␮M furosemide and the lack of potentiation by 0.5 1998) which lack ␦ subunits (Nusser et al., 1998) shows ␮M diazepam of the tonic and spillover currents that we high sensitivity to neurosteroids (Figure 2K). These data see implies that receptors containing ␣6 (but not ␣1) would suggest that the receptors mediating the tonic subunits generate these currents. and spillover IPSC currents contain ␣6 and ␦ subunits. Tonic and Spillover Inhibition 631

Figure 6. Blocking ␣6 Subunit-Containing

GABAA Receptors Potentiates Signal Trans- mission through the Cerebellar Cortex (A) Response of a Purkinje cell (resting potential Ϫ71 mV) to mossy fiber stimulation at 20 Hz in control solution, in 100 ␮M furosemide, and after washing out the furosemide

(ECl ϭϪ61 mV). Small, brief, vertical deflections (Ͻ20 mV) are stimulus artifacts; large deflections are action potentials. (B) As (A), but for 50 Hz stimulation in control solution and furosemide; right panel shows block of response by 20 ␮M NBQX and 50 ␮M AP5. (C) EPSP amplitude (relative to the resting potential) for the first three EPSPs in a stimulus train as a function of mossy fiber (MF) stimulus frequency. (D) Fractional potentiation of the second and third EPSPs in a train by furosemide (first EPSP was too variable to measure potentiation accurately). (E) Time to first action potential produced by a stimulus train. (F) Purkinje cell firing frequency as a function of stimulation frequency. (G) Furosemide (100 ␮M), but not bumetanide (220 ␮M), increases EPSP size and firing rate of a Purkinje cell (resting potential Ϫ74 mV) in response to MF stimulation at 50 Hz. Right panel shows block of response by glutamate receptor blockers as in (B). (H) In coronal slices, furosemide (100 ␮M) potentiated the Purkinje cell response to mossy fiber stimulation (at 100 Hz), as in parasagittal slices (D–F): the EPSP was increased in size, the time to the first action potential was reduced, and the firing frequency was increased (seven cells). (I) In coronal slices furosemide (100 ␮M) had no effect on the Purkinje cell response to parallel fiber stimulation (100 Hz, five cells).

Going against the idea that ␦ subunits suppress the balance the evidence available favors the idea that neu- sensitivity of ␣6-containing receptors to neurosteroids, rosteroid insensitivity in ␣6-containing receptors implies Wohlfarth et al. (2002) found no difference in the sensitiv- the presence of ␦ subunits and, hence, that both the ity to THDOC between ␣6 subunit-containing receptors tonic current and the spillover component of the IPSC containing ␦ or ␥2L. However, as Wohlfarth et al. (2002) are mediated by receptors containing ␣6 and ␦ subunits. discuss, this contradicts in vivo data showing that when These receptors are located exclusively extrasynapti- more ␦ subunit is expressed in older animals, there is cally (Nusser et al., 1998). less sensitivity to neurosteroids (Figures 2C and 2F; Zhu Two aspects of the ␣6␦ subunit combination suggest et al., 1996; Zhu and Vicini, 1997), and differences in the that these receptors are specialized to generate tonic response to neurosteroids between recombinant and in inhibition by responding to persistent low background vivo systems may reflect different phosphorylation of levels of GABA. First, this combination produces the receptors (Fancsik et al., 2000). Finally, a mouse lacking lowest EC50 for activation by GABA (0.2 ␮M for ␣6␤2␦

␦ subunits (Mihalek et al., 1999) shows less sensitivity compared with 2 ␮M for ␣6␤2␥2 and 10 ␮M for ␣1␤2␥2; to neurosteroids on behavioral measures, but these Saxena and MacDonald, 1996). This EC50 is low enough measures almost certainly do not simply reflect the to lead to tonic receptor activation by the resting extra- properties of the ␣6 subunit-containing receptors with cellular level of GABA: even in the absence of GABA which we are concerned, and the decreased sensitivity release, GABA transporters can only lower [GABA]o to could easily be explained by developmental changes around 0.4 ␮M (Attwell et al., 1993). Second, GABAA secondary to the ␦ knockout since the total number of receptors containing the ␣6 and ␦ subunits do not readily

GABAA receptors in the whole brain (even in regions desensitize (Saxena and MacDonald, 1994; Tia et al., where ␦ is not expressed) was halved in the knockout. 1996b). Although the issue is apparently not yet resolved, on Consistent with our conclusion that the tonic inhibi- Neuron 632

tory conductance is mediated by receptors containing 1969; Tyrrell and Willshaw, 1992). This suggests that

␣6 subunits, the tonic conductance is absent in mice tonic and spillover inhibition of granule cells may be an lacking the ␣6 subunit (Brickley et al., 2001). However, important determinant of the animals’ motor repertoire. knocking out ␣6 has complicated secondary effects. First, it also downregulates GABA receptor ␦, ␤ , ␤ , ␥ , A 2 3 2 Experimental Procedures and ␣1 subunit expression in granule cells (Nusser et al., 1999), so, from the knockout data alone, receptors Patch-Clamp Recording containing ␣1 subunits, rather than ␣6, might generate Recording from cells in thin (160–250 ␮m) slices of cerebellum from some or all of the tonic conductance. Second, the 35- to 45-day-old Sprague-Dawley rats (Rossi and Hamann, 1998) knockout upregulates Kϩ channels (Brickley et al., 2001), was at 29ЊϮ3ЊC, in solution containing: 126 mM NaCl, 24 mM the hyperpolarizing effect of which might reduce GABA NaHCO3, 1 mM NaH2PO4, 2.5 mM KCl, 2.5 mM CaCl2,10mM D-glucose, 2 mM MgCl (gassed with 95% O /5% CO ), pH 7.4. release or promote GABA uptake and so abolish the 2 2 2 Slices were parasagittal except for the experiments on coronal slices tonic conductance. Third, it downregulates GABAA re- in Figures 6H and 6I. Drugs were dissolved directly into extracellular ceptor ␣1 and ␤2 subunits in the forebrain (Uusi-Oukari solution or from a stock solution in DMSO (Ͻ0.1% final concentra- et al., 2000), which may lead to altered action potential tion) and applied by bath perfusion. For voltage clamping, pipettes firing throughout the brain and compensatory changes were filled with solution containing: 135 mM CsCl, 4 mM NaCl, 0.5 in the cerebellum. Finally, given its effects on the expres- mM CaCl2, 10 mM HEPES, 5 mM EGTA, 2 mM MgATP, 0.5 mM ϩ NaGTP, 10 mM QX-314 (to suppress voltage-gated sodium current), sion of GABAA and K channel proteins, the knockout pH set to 7.2 with CsOH (ECl ϭ 0 mV). For whole-cell, current-clamp may change the expression of other as yet unstudied recording, pipettes contained 132.3 mM K-gluconate, 7.7 mM KCl, proteins that modulate GABA receptors or release. Con- 4 mM NaCl, 0.5 mM CaCl2, 10 mM HEPES, 5 mM EGTA, 2 mM sequently, it is not possible to attribute the tonic conduc- MgATP, 0.5 mM Na2GTP, pH set to 7.2 with KOH (ECl ϭϪ61 mV), tance to ␣6 subunit-containing receptors solely on the and current injection was used to set the resting potential of granule basis of the knockout data: a pharmacological ap- cells at –65 mV and of Purkinje cells at Ϫ60 to Ϫ75 mV. proach, as we have used, was needed to demonstrate their involvement in generating the tonic conductance. Evoked Responses -M⍀ filled with extracellular medium) or con 2–1ف Since furosemide-sensitive, diazepam- and neuro- Glass (resistance steroid-insensitive GABAA receptors generate both the centric bipolar tungsten stimulating electrodes were placed in the majority (89%) of the charge transfer during the IPSC white matter to activate mossy fibers or in the molecular layer to from Golgi cells (Figure 2L) and all of the tonic inhibitory activate parallel fibers. Trains of stimuli (10–200 ␮s long) were 200– conductance, which is at least 3-fold larger than the 400 ms in duration and delivered only every 30 s to prevent changes in synaptic efficacy (D’Angelo et al., 1995). Receptor blockers were mean conductance activated by synaptically released used to check that action potentials were produced by gluta- GABA (Figure 5G), they have a profound effect on cere- matergic input and not by directly exciting the cell (Figures 5 and bellar information processing. Blocking ␣6 subunit-con- 6). When recording from Purkinje cells, we ensured that climbing taining receptors with furosemide doubled the gain for fibers were not activated by examining the synaptic current evoked the transmission of information from mossy fibers to in voltage-clamp mode. Early and spillover components of the IPSC granule cells (Figure 5D), and the gain change would be were quantified (Rossi and Hamann, 1998) by fitting a sum of two exponentials to the IPSC decay (Figure 2G). Early and spillover ␶ even larger if the block by furosemide were complete. values in 21 cells were 33.5 Ϯ 5.5 ms (similar to P12 animals; Rossi This change of gain may occur partly because inhibition and Hamann, 1998) and 796 Ϯ 41 ms (3-fold longer than in P12 prevents the granule cell EPSP from reaching the range animals), and amplitudes were 31.8 Ϯ 3.8 and 9.8 Ϯ 1.4 pA. Drug where there is less Mg2ϩ-blockage of NMDA receptors effects on these components were measured at 30 ms (after the (D’Angelo et al., 1995) and decreases the membrane rise of the spillover component, which has a 10%–90% rise time in time constant and the duration over which temporal adult cells of 16 ms [Rossi and Hamann, 1998]; the early component summation of synaptic inputs can occur (Bernander et was defined as total current minus the slow exponential) and at 200 ms (spillover component) after the stimulus. Charge transfer of the al., 1991; Gabbiani et al., 1994). In addition to altering spillover component was calculated as the product of its amplitude the gain of information transfer through the cerebellar and decay time constant (this assumes it has zero rise time; allowing granular layer, blocking tonic inhibition is predicted by for its rise time would reduce the derived value by only 1%). Charge computer simulations to increase the tendency of the transfer of the early component was calculated as the total charge granule cell-Golgi cell loop to generate oscillatory firing entry minus the spillover charge entry. for afferent input patterns on the mossy fibers that are more realistic than those evoked by electrical stimula- Statistics tion (Maex and De Schutter, 1998). As a result of the Data are mean Ϯ SEM, and significance was assessed with a two- decreased granule cell excitability produced by tonic tailed paired Student’s t test. and spillover inhibition, the overall gain for information transmission through the cerebellar cortex, from the Acknowledgments mossy fiber input to the Purkinje cell output, is decreased (Figure 6). This work was supported by the Wellcome Trust, a Burroughs- Our data show that tonic and spillover inhibition of Wellcome fellowship to D.J.R., and a Royal Society-Wolfson Re- granule cells, mediated by furosemide-sensitive, diaze- search Merit Award to D.A. We thank Drs. Wohlfarth, Bianchi, and pam- and neurosteroid-insensitive receptors, decreases Macdonald for a preprint of their 2002 paper; Brian Billups for soft- the number of granule cells excited by mossy fiber input. ware; Jeremy Lambert for helpful discussion; and Ce´ line Auger, Computational models of cerebellar cortex have empha- Daniela Billups, Alasdair Gibb, Mike Ha¨ usser, Paı¨kan Marcaggi, and Angus Silver for comments on the manuscript. sized that decreasing the number of granule cells that are simultaneously active increases the number of motor programs that can be stored in the cerebellum (Marr, Received June 14, 2001; revised September 18, 2001. Tonic and Spillover Inhibition 633

References G. (2000). GABAA receptors: immunocytochemical distribution of 13 subunits in the adult rat brain. Neuroscience 101, 815–850. Attwell, D., Barbour, B., and Szatkowski, M. (1993). Nonvesicular Quirk, K., Gillard, N.P., Ragan, C.I., Whiting, P.J., and McKernan, release of neurotransmitter. Neuron 11, 401–407. R.M. (1994). Model of subunit composition of ␥-aminobutyric acid Bernander, O¨ ., Douglas, R.J., Martin, K.A.C., and Koch, C. (1991). A receptor subtypes expressed in rat cerebellum with respect to Synaptic background activity influences spatiotemporal integration their ␣ and ␥/␦ subunits. J. Biol. Chem. 269, 16020–16028. in single pyramidal cells. Proc. Natl. Acad. Sci. USA 88, 11569– Rossi, D.J., and Hamann, M. (1998). Spillover-mediated transmission 11573. at inhibitory synapses promoted by high affinity ␣6 subunit GABAA Brickley, S.G., Cull-Candy, S.G., and Farrant, M. (1996). Develop- receptors and glomerular geometry. Neuron 20, 783–795. ment of a tonic form of synaptic inhibition in rat cerebellar granule Rossi, D.J., and Slater, N.T. (1993). The developmental onset of cells resulting from persistent activation of GABAA receptors. J. NMDA receptor-channel activity during neuronal migration. Neuro- Physiol. 497, 753–759. pharmacology 32, 1239–1248. Brickley, S.G., Revilla, V., Cull-Candy, S.G., Wisden, W., and Farrant, Sah, P., Hestrin, S., and Nicoll, R.A. (1989). Tonic activation of NMDA M. (2001). Adaptive regulation of neuronal excitability by a voltage- receptors by ambient glutamate enhances excitability of neurons. independent potassium conductance. Nature 409, 88–92. Science 246, 815–818. D’Angelo, E., De Filippi, G., Rossi, P., and Taglietti, V. (1995). Synap- Saxena, N.C., and MacDonald, R.L. (1994). Assembly of GABAA re- tic excitation of individual rat cerebellar granule cells in situ: evi- ceptor subunits: role of the ␦ subunit. J. Neurosci. 14, 7077–7086. dence for the role of NMDA receptors. J. Physiol. 484, 397–413. Saxena, N.C., and MacDonald, R.L. (1996). Properties of putative Fancsik, A., Linn, D.M., and Tasker, J.G. (2000). Neurosteroid modu- cerebellar ␥-aminobutyric acidA receptor isoforms. Mol. Pharmacol. lation of GABA IPSCs is phosphorylation dependent. J. Neurosci. 49, 567–579. 20, 3067–3075. Sigel, E., and Baur, R. (2000). Electrophysiological evidence for the Gabbiani, F., Midtgaard, J., and Kno¨ pfel, T. (1994). Synaptic integra- coexistence of ␣1 and ␣6 subunits in a single functional GABAA recep- tion in a model of cerebellar granule cells. J. Neurophysiol. 72, tor. J. Neurochem. 74, 2590–2596. 999–1009. Thompson, S.A., Arden, S.A., Marshall, G., Wingrove, P.B., Whiting, Isaacson, J.S. (1999). Glutamate spillover mediates excitatory trans- P.J., and Wafford, K.A. (1999). Residues in transmembrane domains mission in the rat olfactory bulb. Neuron 23, 377–384. I and II determine GABAA receptor subtype-selective antagonism Jechlinger, M., Pelz, R., Tretter, V., Klausberger, T., and Sieghart, W. by furosemide. Mol. Pharmacol. 55, 993–999. (1998). Subunit composition and quantitative importance of hetero- Tia, S., Wang, J.F., Kotchabhakdi, N., and Vicini, S. (1996a). Develop- ␣ oligomeric receptors: GABAA receptors containing 6 subunits. J. mental changes of inhibitory synaptic currents in cerebellar granule Neurosci. 18, 2449–2457. neurons: role of GABAA receptor ␣6 subunit. J. Neurosci. 16, 3630– Kaneda, M., Farrant, M., and Cull-Candy, S.G. (1995). Whole-cell 3640. and single-channel currents activated by GABA and glycine in gran- Tia, S., Wang, J.F., Kotchabhakdi, N., and Vicini, S. (1996b). Distinct ule cells of the rat cerebellum. J. Physiol. 485, 419–435. deactivation and desensitization kinetics of recombinant GABAA re- Korpi, E.R., and Lu¨ ddens, H. (1997). Furosemide interactions with ceptors. Neuropharmacology 35, 1375–1382. brain GABAA receptors. Br. J. Pharmacol. 120, 741–748. Tyrrell, T., and Willshaw, D. (1992). Cerebellar cortex: its simulation Korpi, E.R., Kuner, T., Seeburg, P.H., and Lu¨ ddens, H. (1995). Selec- and the relevance of Marr’s theory. Philos. Trans. R. Soc. Lond. B tive antagonist for the cerebellar granule cell-specific GABAA recep- Biol. Sci. 336, 239–257. tor. Mol. Pharmacol. 47, 283–289. Uusi-Oukari, M., Heikkila, J., Sinkkonen, S.T., Makela, R., Hauer, B., Laurie, D.J., Seeburg, P.H., and Wisden, W. (1992a). The distribution Homanics, G.E., Sieghart, W., Wisden, W., and Korpi, E.R. (2000). of thirteen GABAA receptor subunit mRNAs in the rat brain II: olfac- Long-range interactions in neuronal gene expression: evidence from tory bulb and cerebellum. J. Neurosci. 12, 1063–1076. gene targeting in the GABAA receptor ␤2-␣6-␣1-␥2 subunit gene clus- Laurie, D.J., Wisden, W., and Seeburg, P.H. (1992b). The distribution ter. Mol. Cell. Neurosci. 16, 34–41. of thirteen GABAA receptor subunit mRNAs in the rat brain III: embry- Wall, M.J., and Usowicz, M.M. (1997). Development of action poten- onic and postnatal development. J. Neurosci. 12, 4151–4172. tial-dependent and independent spontaneous GABAA receptor- Maex, R., and De Schutter, E. (1998). Synchronization of Golgi and mediated currents in granule cells of postnatal rat cerebellum. Eur. granule cell firing in a detailed network model of the cerebellar J. Neurosci. 9, 533–548. granule cell layer. J. Neurophysiol. 80, 2521–2537. Wisden, W., Korpi, E.R., and Bahn, S. (1996). The cerebellum: a Marr, D. (1969). A theory of cerebellar cortex. J. Physiol. 202, model system for studying GABAA receptor diversity. Neuropharma- 437–470. cology 35, 1139–1160. Mihalek, R.M., Banerjee, P.K., Korpi, E.R., Quinlan, J.J., Firestone, Wohlfarth, K.M., Bianchi, M.T., and Macdonald, R.L. (2002). En- L.L., Mi, Z.P., Lagenaur, C., Tretter, V., Sieghart, W., Anagnostaras, hanced neurosteroid potentiation of ternary GABAA receptors con- S.G., et al. (1999). Attenuated sensitivity to neuroactive steroids in taining the delta subunit. J. Neurosci., in press.

GABAA receptor delta subunit knockout mice. Proc. Natl. Acad. Sci. Zhu, W.J., and Vicini, S. (1997). Neurosteroid prolongs GABAA chan- USA 96, 12905–12910. nel deactivation by altering kinetics of desensitized states. J. Neu- Nusser, Z., Sieghart, W., and Somogyi, P. (1998). Segregation of rosci. 17, 4022–4031. different GABAA receptors to synaptic and extrasynaptic mem- Zhu, W.J., Wang, J.F., Krueger, K.E., and Vicini, S. (1996). ␦ subunit branes of cerebellar granule cells. J. Neurosci. 18, 1693–1703. inhibits neurosteroid modulation of GABAA receptors. J. Neurosci. Nusser, Z., Ahmad, Z., Tretter, V., Fuchs, K., Wisden, W., Sieghart, 16, 6648–6656.

W., and Somogyi, P. (1999). Alterations in the expression of GABAA receptor subunits in cerebellar granule cells after the disruption of the ␣6 subunit gene. Eur. J. Neurosci. 11, 1685–1697. Otis, T.S., and Mody, I. (1992). Modulation of decay kinetics and frequency of GABAA receptor-mediated spontaneous inhibitory postsynaptic currents in hippocampal neurons. Neuroscience 49, 13–32. Payne, J.A. (1997). Functional characterization of the neuronal spe- ϩ cific K-Cl cotransporter: implications for [K ]o regulation. Am. J. Physiol. 273, C1516–C1525. Pirker, S., Schwarzer, C., Wieselthaler, A., Sieghart, W., and Sperk,